Uploaded by GUNAY CALISKAN

The sensitivity analysis of even order biquadratic

advertisement
See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/238600954
The sensitivity analysis of even order biquadratic elliptic filters
Article in Bulletin of the Polish Academy of Sciences, Technical Sciences · March 2006
CITATION
READS
1
70
2 authors, including:
Tomasz Adrikowski
Silesian University of Technology
15 PUBLICATIONS 28 CITATIONS
SEE PROFILE
All content following this page was uploaded by Tomasz Adrikowski on 31 October 2016.
The user has requested enhancement of the downloaded file.
BULLETIN OF THE POLISH ACADEMY OF SCIENCES
TECHNICAL SCIENCES
Vol. 54, No. 1, 2006
The sensitivity analysis of even order biquadratic elliptic filters
M. PASKO∗ and T. ADRIKOWSKI
Institute of Theoretical and Industrial Electrotechnics, Silesian University of Technology, 16 Akademicka St., 44-100 Gliwice, Poland
Abstract. In this paper, the sensitivity analysis of the elliptic filters realized by using biquadratic structures was carried out. The influence
of spread the structure parameter values on the shape of the frequency characteristic of the filter transmittance modulus was analyzed. The
analysis was limited to the case of even order low-pass filter. Defining the proper class of the sensitivity coefficients, the changes influence
of individual structure parameters on the deviation of basic parameter values of the characteristic was considered. The considerations were
illustrated by the numerical example.
Key words: elliptic filter, biquadratic structures, frequency filter characteristic, sensitivity analysis, sensitivity coefficients.
1. Introduction
the nominal values
The elliptic filter from among all known types of filters distinguishes the largest selectivity [1]. Frequency characteristics of
transmittance modulus have oscillations in the pass-band and
the stop-band. Between those bands, it is possible to distinguish the transition-band in which there is a sudden fall of filter transmission. The characteristic of the example low-pass
elliptic filter of order r = 4 is shown in Fig. 1.
¯
2
Y
¯
−ω 2 + ωz2in
¯
|K(jω)|n = Ku0n
¯ −ω 2 + j2σin ω + ω 2
r
pin
i=1
¯
¯
¯,
¯
(1)
where
– σin , ωpin , ωzin – the nominal parameter values of i-th biquadratic structure,
– Ku0n – the nominal value of the fixed factor.
In Fig. 2 general course of the nominal characteristic of lowpass elliptic filter of even order r is presented. The characteristic is clear-out described by the given nominal parameters:
– Kurn – the nominal maximum value of transmittance modulus in the pass-band,
– Kupn – the nominal minimum value of transmittance modulus in the pass-band,
– ωgrn – the nominal limited pulsation, above that transmittance modulus falls below Kupn value,
– Kusn – the nominal maximum value of transmittance modulus in the stop-band,
– ωsn – the nominal pulsation, below that transmittance modulus grows above the maximum value of the stop-band
equals Kusn .
Fig. 1. The characteristic of the example low-pass elliptic filter of
order r = 4
In this paper, the proposed sensitivity analysis of elliptic
filters is limited to the case of low-pass filter of even order [2].
A transmittance of even order r filter can be expressed by the
product of 2r biquadratic factors and by the resultant fixed factor. It is possible to build elliptic filters in the class of active
devices with the use of biquadratic structures that realize the
individual biquadratic factors [3–5]. The fixed factor can be
attainable outside biquadratic structures in the single proportional structure.
In the pass-band (0 < ω < ωgrn ) transmittance modulus has oscillation course and oscillates between the minimum
value Kupn and the maximum value Kurn . The level Kurn
is achieved by all j-th local maximums Kurj which are in the
pass-band.
r
(2)
Kurn = Kurj , j = 1, 2, ..., .
2
The nominal value Kupn equals to Kup1 value of transmittance modulus, that is achieved for ω = 0
Kupn = Kup1 = lim |Ku (jω)|n
ω→0
2. The nominal characteristic
The nominal frequency characteristic of transmittance modulus |K(jω)|n is attainable, when all structure parameters adopt
∗ e-mail:
and all j-th local minimums Kupj of the pass-band
r
Kupn = Kupj , j = 2, 3, ..., .
2
marian.pasko@polsl.pl
101
(3)
(4)
M. Pasko and T. Adrikowski
Fig. 2. The nominal characteristic of the low-pass elliptic filter of
even order r = 4
Fig. 3. The deviation characteristic of even order r low-pass elliptic
filter
σi = σin (1 + δσi ) ,
The measure of oscillations in the pass-band is the passband ripple expressed in dB
µ
¶
Kurn
R(dB)n = 20 log
.
(5)
Kupn
The maximum value Kusn of the stop-band equals to Kus1
value of transmittance modulus, that is achieved for ω = ∞
Kusn = Kus1 = lim |Ku (jω)|n
(6)
ω→∞
and all j-th local maximums Kusj of the band
r
Kusn = Kusj , j = 2, 3, ..., .
2
(7)
The transition-band included between the pulsation ωgrn
and ωsn characterized a sudden fall of transmittance modulus
value. The measure of this fall is an average slope of the characteristic NdB/okt expressed in dB/okt, described according
to the formula
³
´
K
20 log Kupn
usn
³
´ .
NdB/oktn =
(8)
ω
log2 ωgrn
sn
ωpi = ωpin (1 + δωpi ) ,
ωzi = ωzin (1 + δωzi ) ,
with
δσi , δωpi , δωzi – relative deviation of the parameter values:
σi , ωpi , ωzi from the nominal values: σin , ωpin , ωzin , Ku0 –
the real value of the fixed factor.
In Fig. 3 there is presented a general course of the deviation characteristic of even order r low-pass elliptic filter.
The parameters describing characteristic gave in deviation
from the nominal values: Kurn , Kupn , ωgrn , Kusn , ωsn to
the values: Kur , Kup , ωgr , Kus , ωs . They are a bit differently defined than for the nominal case.
In the pass-band (0 < ω < ωgr ) j-th values: Kurj and
Kupj aren’t equal, in that case it is necessary to accept for the
parameter Kur the biggest value from among j-th Kurj values
r
©
ª
2
Kur = max Kurj ,
j=1
however for Kup parameter it is necessary to accept the smallest value from among j-th Kupj values
r
©
ª
2
Kup = min Kupj .
j=1
3. The deviation characteristic
In practice, real values of structure parameters differ from
nominal values, for the sake of spread of the element values, from these elements the structures are built. Deviation
from the nominal value even of one parameter of the structure causes characteristic deviation from nominal case. A
frequency deviation characteristic of transmittance modulus
|K(jω)| is achieved in the case when all structure parameters
adopt real values
¯
2
Y
¯
−ω 2 + ωz2i
¯
|K(jω)| = Ku0
¯ −ω 2 + j2σi ω + ω 2
r
i=1
pi
¯
¯
¯,
¯
(9)
where
σi , ωpi , ωzi – the real parameter values of the i-th biquadratic
structure
102
(10)
(11)
The limited pulsation ωgr is the one above that transmittance modulus falls below the Kup value, however real passband ripple expressed in dB is characterized by the expression
µ
¶
Kur
R(dB) = 20 log
.
(12)
Kup
In the stop-band (ωsn < ω < ∞) j-th values Kusj aren’t
equal, so in this case it is necessary to accept for the parameter
Kus the biggest value from among j-th Kusj values
r
©
ª
2
Kus = max Kusj .
j=1
(13)
The pulsation ωs is the one below that transmittance modulus grows over the Kus value, however the average slope of
the deviation characteristic NdB/okt in the transition-band expressed in dB/okt, can be described according to the formula
Bull. Pol. Ac.: Tech. 54(1) 2006
The sensitivity analysis of even order biquadratic elliptic filters
NdB/okt
³
´
K
20 log Kup
us
³
´ .
=
ωgr
log2 ωs
(14)
4. The coefficients describing sensitivity
of characteristic to changes of structure
parameters
Determining proper class of sensitivity coefficients it is possible to count relative change of each characteristic parameters
for any combinations relative deviations of structure parameters. The essential class of the sensitivity coefficients must determine the sensitivity very detailed characteristic parameters
to changes of each structure
Structure
¡ parameter separately.
¢
parameters: σi , ωpi , ωzi i = 1, 2, ..., 2r create the vector of
influential parameters Y about 3r
2 elements
h
iT
Y = Y1 , Y2 , ..., Yk , ..., Y 3r2
·
¸T
= σ1 , σ2 , ..., σ r2 , ωp1 , ωp2 , ..., ωp r , ωz1 , ωz2 , ..., ωz r
,
2
for k = 1, 2, ...
2
3r
.
2
(15)
In Fig. 4 there is showed general characteristic of filter, for
deviation from nominal value k-th structure parameter Yk =
Ykn + ∆Yk . The other structure parameters accept the nominal values. In this figure there are marked all detailed characteristic parameters that sensitivity should be described for the
changes of each k-th structure parameter Yk . There are all j-th
values of the following parameters: Kurj , Kupj , ωgrj , Kusj ,
ωsj (j = 1, 2, ..., 2r ). The meaning of parameters: Kurj , Kupj ,
Kusj has been explained by describing the parameters of the
deviation characteristic (look at point 3).
It is necessary to explain the meaning of the parameters:
ωgrj , ωsj
– j-th value of the parameter ωgrj is the pulsation, that
|K(jω)| achieves j-th value Kupj ,
– j-th value of the parameter ωsj is the pulsation, that
|K(jω)| achieves j-th value Kusj .
Within the confines of the sensitivity analysis it is necessary to determine the sensitivity coefficient of all j-th values
of the characteristic parameters: Kurj , Kupj , ωgrj , Kusj , ωsj
for the changes of each k-th structure parameter Yk
Kur
Kup
ωgr
Kus
ωs
SYk j , SYk j , SYk j , SYk j , SYk j ,
k = 1, 2, ...,
r
3r
, j = 1, 2, ..., .
2
2
Kurj
Kus
ωs
Kupj
, SYk
Kurj
SYk
ωgrj
SYk
ωs
=
Kurj − Kurn Yk
Kupj − Kupn Yk
Kup
, SYk j =
,
Kurn
∆Yk
Kupn
∆Yk
=
Kusj − Kusn Yk
ωgrj − ωgrn Yk
Kus
, SYk j =
,
ωgrn
∆Yk
Kus n
∆Yk
SYk j =
ωsj − ωsn Yk
3r
r
, k = 1, 2, ..., , j = 1, 2, ...,
ωsn
∆Yk
2
2
(17)
5. Calculation of characteristic deviations
On the basis of sensitivity coefficients (17) it is possible to
count the relative change of each characteristic parameter, for
any deviation combination δYk (k = 1, 2, ..., 3r
2 ) of particular
structure parameters, according to the formulas
 3r

2 ³
X
´
r
K
ur
2
δKur = max
SYk j δYk
,
(18)
j=1 

k=1
r
2
δKup = min
 3r
2 ³
X

=
2 ³
X
Kupj
SYk
k=1
3r
,
SYk j , SYk j , SYk j are possible to determine in the numerical way for each k-th structure parameter separately. In order
to achieve this aim it is necessary to deviate the value of k-th
structure parameter Yk about any not large known value ∆Yk ,
keeping the nominal values of the rest structure parameters.
Bull. Pol. Ac.: Tech. 54(1) 2006
For such deviated k-th structure parameter it is necessary
to determine, all j-th values of the structure parameters: Kurj ,
Kupj , ωgrj , Kusj , ωsj , suitable for this deviation, taking them
back to the nominal values: Kurn , Kupn , ωgrn , Kusn , ωsn .
The coefficient values for each k-th deviated structure parameter can be counted according to the formulas
j=1
The individual k-th coefficient values: SYk
ωgr
(16)
Fig. 4. General characteristic of low-pass elliptic filter for deviated
from nominal value of k-th structure parameter Yk = Yk n + ∆Yk .
The other structure parameters accept the nominal values
K

´
δYk

(19)
´
SYkupw δYk , 1 6 w 6
k=1
3r
δωgr =
2
X
¡
¢
ω
SYkgrw δYk ,
r
,
2
(20)
k=1
103
M. Pasko and T. Adrikowski
δKus

 3r
2 ³
X
´
Kusj
= max
SYk δYk
j=1 

r
2
k=1
(21)
3r
2
´
X ³ Kus
r
=
SYk q δYk , 1 6 q 6 ,
2
Kurj
H SYk
(k = 1, 2...6, j = 1, 2)
K
K
K
K
SY1ur1 = −8.5 · 10−1 , SY2ur1 = −6.6 · 10−3 ,
k=1
3r
2 ³
´
X
ωs
δωs =
SYk q δYk .
The sensitivity coefficients are determined by the formulas
(17), deviating in turn the values of each structure parameters
about +5% (∆Yk = 0.05) relative to the nominal value
(22)
K
SY3ur1 = −6.7 · 10−1 , SY4ur1 = −2.2, SY5ur1 = 2.1,
K
K
K
SY6ur1 = 2.1, SY1ur2 = −2.8 · 10−1 , SY2ur2 = −8.6 · 10−1 ,
k=1
The sensitivity coefficients (17) on the assumption that
the changes of characteristic parameters evoked by changes of
structure parameters are linear. In this connection, these presented formulas to relative deviations of characteristic parameters (18), (19), (20), (21), (22) can be applied when the relative
deviations of characteristic structures are not large (do not exceed 10% of the nominal value), which, however, is fulfilled in
practice.
On the basis of the relative deviations of characteristic parameters it is possible to count relative values of these parameters
K
K
K
SY3ur2 = 5.2 · 10−1 , SY4ur2 = −3.5, SY5ur2 = 2.1.
K
SY6ur2 = 2.5,
Kupj
H SYk
(k = 1, 2...6, j = 1, 2)
K
K
K
SY1up1 = 2.8 · 10−4 , SY2up1 = 2.8 · 10−4 , SY3up1 = −1.9,
K
K
K
SY4up1 = −1.9, SY5up1 = 2.1, SY6up1 = 2.1,
K
K
K
K
SY1up2 = 4.9 · 10−1 , SY2up2 = −8.0 · 10−2
K
Kur = Kurn (1 + δKur ) , Kup = Kupn (1 + δKup ) ,
SY3up2 = 7.4 · 10−1 , SY4up2 = −3.9, SY5up2 = 2.1
ωgr = ωgrn (1 + δωgr ) , Kus = Kusn (1 + δKus )
SY6up2 = 2.3.
(23)
K
ωgrj
H SYk
ωs = ωsn (1 + δωs ) .
Example. To illustrate the presented considerations in this example, the sensitivity analysis of normalized (ωgrn = 1) elliptic filter order r = 4 with pass-band ripple equals 3 dB was
carried out. Transmittance modulus of this filter can be presented by a product of transmittance modulus and two component biquadratic structures
¯
¯
¯
¯
−ω 2 + ωz21 n
¯
|K(jω)|n = Ku0 ¯¯
2
2
−ω + j2σ1 n ω + ωp1 n ¯
(24)
¯
¯
¯
¯
−ω 2 + ωz22 n
¯
¯
,
ׯ
−ω 2 + j2σ2 n ω + ω 2 ¯
p2 n
ω
SY1gr1
ω
SY3gr1
ω
SY5gr1
ω
SY1gr2
ω
SY3gr2
ω
SY5gr2
(k = 1, 2...6, j = 1, 2)
ω
= −2.6 · 10−2 , SY2gr1 = −8.7 · 10−2 ,
ω
= 2.5 · 10−1 , SY4gr1 = 7.7 · 10−1 ,
ω
= 8.7 · 10−3 , SY6gr1 = 5.5 · 10−2 ,
ω
= 2.9 · 10−2 , SY2gr2 = −7.7 · 10−2 ,
ω
= −2.9 · 10−2 , SY4gr2 = 1.0
ω
= 4.2 · 10−3 , SY6gr2 = 2.9 · 10−2 .
Kusj
H SYk
(k = 1, 2...6, j = 1, 2)
K
K
SY1us1 = −2.8 · 10−2 , SY2us1 = −4.6 · 10−3 ,
where
K
σ1n = 0.21396, ωp1n = 0.47121, ωz1n = 4.9221,
K
K
K
K
K
K
ωs
H SYk j (k = 1, 2...6, j = 1, 2)
Structure parameters create six-elemental vector of influential parameters Y
T
T
= [σ1n , σ2n , ωp1n , ωp2n , ωz1n , ωz2n ] .
(25)
A course of nominal characteristic is compatible with given
in Fig. 1. The nominal values of characteristic parameters are
equal to
Kurn = 1.0000, Kupn = 0.70710, ωgrn = 1.0000,
Kusn = 1.2936 · 10−3 , ωsn = 2.0000.
104
K
SY3us2 = 0.0, SY4us2 = 0.0, SY5us2 = 0.0, SY6us2 = 0.0
ωp2 ωp2
1
= 21n 22n √ .
ωz1n ωz2n 2
Y = [Y1 , Y2 , Y3 , Y4 , Y5 , Y6 ]
K
SY6us1 = −2.9, SY1us2 = 0.0, SY2us2 = 0.0
σ2n = 0.075257, ωp2n = 0.95752, ωz2n = 2.1432,
Ku0n
K
SY3us1 = 6.0 · 10−2 , SY4us1 = 2.9 · 10−1 , SY5us1 = 3.0,
(26)
ω
ω
SY1s1 = −1.1 · 10−3 , SY2s1 = −2.0 · 10−4 ,
ω
ω
SY3s1 = 3.0 · 10−3 , SY4s1 = 1.8 · 10−2 ,
ω
ω
ω
ω
SY5s1 = −2.5 · 10−2 , SY6s1 = 1.0,
SY1s2 = −2.7 · 10−3 , SY2s2 = −5.0 · 10−4 ,
ω
ω
ω
ω
SY3s2 = 6.2 · 10−3 , SY4s2 = 3.3 · 10−2 ,
SY5s2 = 1.2 · 10−1 , SY6s2 = 8.3 · 10−1 .
Deviations of characteristic parameters are determined as
an example deviation combination of structure parameters
Bull. Pol. Ac.: Tech. 54(1) 2006
The sensitivity analysis of even order biquadratic elliptic filters
View publication stats
δσ1 = −0.01, δσ2 = −0.01, δωp1 = +0.01,
δωp2 = −0.01, δωz1 = +0.01, δωz2 = +0.01.
They are equal to
(
2
δKur = max
j=1
(27)
= max
½
6 ³
X
Kurj
SYk
6 ³
X
´
= 9.24 · 10
k=1
)
= 9.76 · 10−2 ∼
= +9.8 · 10−2 ,
6 ³
´
X
Kup
= min
SYk j δYk
)
2
Kur = Kurn (1 + δKur ) = 1.096,
(33)
Kup = Kupn (1 + δKup ) = 7.361 · 10−1 ,
(34)
ωgr = δωgrn (1 + δωgr ) = 9.966 · 10−1 ,
(35)
Kus = δKusn (1 + δKus ) = 1.294 · 10−3 ,
(36)
ωs = δωsn (1 + δωs ) = 2.019.
(37)
k=1
(
)
6 ³
6 ³
´ X
´
X
Kup1
Kup2
= min
SYk δYk ,
SYk δYk
(29)
k=1
k=1
½
¾
= min 4.10 · 10−2 , 9.65 · 10−2
−−(w=1)−−
= 4.10 · 10−2 ∼
= +4.1 · 10−2 , w = 1,
δωgr =
6
X
¡
6
¢ X
¡ ωgr1
¢
ω
SYkgrw δYk =
SYk δYk
k=1
= −3.39 · 10
∼
= −3.4 · 10−3 ,
(
δKus
(30)
k=1
−3
6 ³
´
X
Kus
SYk j δYk
= max
)
2
j=1
k=1
(
6 ³
6 ³
´ X
´
X
Kus1
K
SYk δYk ,
SYkus2 δYk
= max
½
k=1
k=1
−3
= max −1.56 · 10
Bull. Pol. Ac.: Tech. 54(1) 2006
,
.
(28)
¾
, 9.76 · 10−2
−−−−−−−
(
∼
= +9.2 · 10
−3
On the basis of the formulas (23) it is possible to determine
real values of characteristic parameters
6 ³
´
´ X
K
SYkur2 δYk
,
k=1
(32)
k=1
)
δYk
K
SYkur1 δYk
= max 6.58 · 10
j=1
6
´ X
¡ ωs2
¢
ωs
SYk q δYk =
SYk δYk
−3
−2
δKup
6 ³
X
k=1
k=1
(
δωs =
0.0
)
¾
−(q=2)−
= 0, q = 2,
(31)
R EFERENCES
[1] W.K Chen, The Circuits and Filters Handbook, IEEE Press,
New York, 1995.
[2] L. Matejicek and K. Vrba, “Sensitivity analysis of higher-order
filters”, Electronics Letters 37 (23), 1–5 (2001).
[3] M. Pasko and T. Adrikowski, “Even order elliptic filter design
using OTA-C or OTA-RC biquadratic structures”, IC-SPETO
2003 Gliwice-Niedzica II, 225–228 (2003), (in Polish).
[4] M. Pasko and T. Adrikowski, “Standardization of elliptic filters” SCI Letters “Elektryka” 182, 89–100 (2002), (in Polish).
[5] M. Pasko and T. Adrikowski, “Practical realization of evenorder elliptic filters by using biquadratic structures”, SCI Letters
“Elektryka” 182, 101–122 (2002), (in Polish).
[6] L. Thede, Analog And Digital Filter Design Using C, Prentice
Hall PTR, New Jersey, 1996.
[7] F. Vallette and G.Vasilescu, “A fully recursive approach to the
computation of higher order sensitivities of linear active circuits”, IEEE Transactions on Circuit and Systems 46 (8) 907–
919 (1999).
105
Download